Everything that’s yours, in order.
A red build on one branch. Two reviews waiting on another. Six bot comments on the pull request you thought was finished, and a conflict that appeared overnight on the one that was ready to land. All of it is yours, and none of it tells you which to do first. Limn ranks the lot across every repository you work in — and lets you finish the top of it without leaving the page.
Everything in the next six sections is free, with no repository limit. Run it on your own machine with npx pierre-review and it keeps no credentials at all.
Your day arrives out of order.
At any moment five different jobs want five different things from you — a build to fix, a review to answer, a thread to reply to, a conflict to clear, one sitting green and ready to land. They are all open, they all have your name on them, and nothing puts them in a sequence.
Spread that across six repositories — a reviewer on some, an author on others, a maintainer on a couple you rarely open — and layer three or four review bots reading every pull request on top. The volume is not the problem on its own. The problem is that none of it arrives sorted by how much it has to do with you, today.
GitHub will tell you that something happened. It will not tell you what to do first.
- A build went red on the branch you pushed before you logged off.
- Someone answered a thread on a pull request you had already moved on from.
- Two reviews are waiting on you in a repository you only maintain.
- One of yours is ready to land and has been green since yesterday.
- And a conflict appeared overnight on the one you most wanted to finish.
Free, always · everything in sections 01 to 06 is in the open-core tier.
One queue, and the next thing at the top of it.
Review requests, threads somebody answered, red builds on your branches, conflicts you can clear, pull requests sitting green and ready to land — five different kinds of job, from every repository, in one list with the most actionable at the top.
The ordering is the point, and it is code rather than a model. It weighs how close a job is to finished — an approved pull request outranks one still waiting on a review, a two-file change outranks a forty-file one, and a conflict or a wall of unanswered threads pushes a job down rather than up — then how long it has been sitting, then how directly it is tied to you. The rank is free on every tier; only the sentence explaining each row is Pro.
A card exists only while you genuinely owe an action and have not taken it. There is nothing to tick off and nothing to dismiss, because the list is worked out fresh on every read rather than stored: act, and the card leaves on its own.
The labels are careful about what they claim. Your turn means the work is tied to you. In your repos means somebody opened something on ground you maintain — orbit, not ownership. A card will not tell you a stranger’s pull request is yours.
09:04 · sixty-one items across five repositories, and the first three are the ones you can finish before standup.
Every review thread already triaged.
Each review thread carries a state worked out from the repository itself: resolved, replied but still open, untouched — or likely addressed, which means a commit touched that file after the comment landed.
That last one is a heuristic and the interface says so wherever it appears. A rename or an unrelated edit can trip it. It is offered as a shortcut through forty bot threads, never as a verdict.
Bot comments also arrive already graded for severity, by a classifier trained for this one job — free, and never the vendor’s own badge, which is stored to be shown and not to be believed.
09:11 · fifteen threads on one pull request; three still need you, and one button clears the rest.
Reply, approve, merge — from the board.
One verdict answers “can this land?” and the action sits on the same row. Merge, or arm merge when ready and it lands the moment checks go green — pinned to the exact commit you looked at, so a new push disarms it rather than merging code you never saw.
When GitHub refuses and will not say why, the pull request carries a ranked list of what is actually holding it, each line marked proven or inferred — because only some of it is knowable from what GitHub returns, and pretending otherwise is how a tool loses your trust.
Conflicts are resolved in the app, hunk by hunk, with no worktree and no free typing anywhere in the flow. Running locally, that is free too.
The header carries how far the change reaches — low, medium or high, with the reasons written out. It refuses rather than guesses: where GitHub truncated the file list it will assert “high” but never “low”, because a truncated list is always a floor.
09:19 · two merged, one armed for when the build finishes, one conflict cleared.
And the bot comments, graded before you read them.
Every bot-written comment is graded for severity — nit, minor, major, critical — by a purpose-trained ModernBERT classifier, served as an int8 ONNX artifact. It runs on the CPU you already have: no GPU, no API call, no comment leaving the machine, and nothing metered. Its eight categories come from a separate deterministic pass, so a category is reproducible rather than sampled.
It is measured, not asserted. We hid every label a batch of real bot comments already carried and had them graded from scratch by a human expert. The model picked the same severity seven times in ten — which is exactly how often two human experts agreed with each other on the same comments, so it is reading them about as well as the job allows. CodeRabbit’s own badge matched the expert fewer than five times in ten.
Which is why the vendor’s badge is stored and shown beside ours and never used as an input. Tuning towards agreement with it measurably moves the model away from ground truth, so the two are drawn apart rather than reconciled.
Where two bots flagged the same lines, the pull request says so, so you read the finding once.
Nothing auto-acts · on a grade. Critical is under-recalled, so the product buckets major and critical together as “high” and leaves the decision with you.
And the rest of it, still without paying.
The stream
One chronological feed across every repository in the workspace, with automated accounts left out until you ask for them. Lenses for red builds and for the kinds of pull-request event you care about.
Search and the board
Full-text search over titles, bodies and comments in every repository you track, and a timeline that plots a fortnight of work per person per repository so a catch-up is one picture.
Quiet by default
Notifications fire only for work that is personally yours, and a repository you do not want claiming your turn can be muted without changing what anybody else sees.
The reading, when the list gets long.
Pro does not unlock an action. Everything you can do to a pull request you can already do for nothing, and that is deliberate — a tool that held back the merge button would be holding your work hostage.
What it adds is writing. A short summary at the top of a long pull request. A read on what a red build actually broke, rather than a link to a log. A reason on each of the top items in your worklist and a line saying what can wait. A check on whether a review comment was really addressed, judged from the commits since the thread started rather than from whether somebody clicked resolve.
On a quiet week you will not miss it. On the week you come back to two hundred threads, it is the difference between reading them and triaging them.
Free ranks the list. Pro writes the reason on it.
Free answers one pull request. Pro answers the fortnight.
Limn is open core. The free tier is the product, not a trial — it has no repository limit, no user limit and no expiry, and every action you take on a pull request lives in it. Pro is what happens when you need the same truth one grain up: across repositories, across people, across a period.
What is waiting on me?
The whole worklist, ranked. My Turn, review requests, red builds, conflicts — across every repository.
Same list, with a written reason on each of the top items and a line saying what can safely wait.
What happened on this pull request?
Every thread triaged — untouched, replied, likely addressed, resolved. Blast radius, blockers, merge verdict.
A short written summary, a read on what a red build means, and a check on whether a comment was really addressed.
Can I act without leaving?
Reply, resolve, approve, merge, merge-when-ready, update from trunk, resolve conflicts hunk by hunk.
Everything free does. Acting on a pull request is never the paid part.
Which review bots are worth their seat?
Detect every bot, name it, set its role and record what it costs — and an independent severity grade on every comment it leaves.
The per-bot scoreboard: keep / tune / noisy, dollars per comment your team acted on, over-calling as a weekly line.
How does that compare to other teams?
—
Your bots placed against a 2,204-repository cohort, with a written refusal wherever the cohort is too thin to answer.
How is the team doing?
Flow metrics for the workspace, activity and reach by repository, open pull requests and trunk health.
A stored report per sprint or month, compared like-for-like against the last one, with the coverage stated.
Is a repository about to bite us?
Default-branch health across every repository, the commit that broke it, and the pull request that landed it.
Everything free does. Trunk health is never the paid part — it is the thing you need at the worst moment.
Where is the time actually going?
—
Chronology: every open hour attributed to a reviewer, an author, or waiting to land. Nobody is named.
What do I take into a one-to-one?
—
A report with a section per person you pick. Preparation, not a ranking — there is no leaderboard anywhere in it.
Can I keep the noise down?
Bots hidden by default, notifications only for work that is personally yours, and a per-repository mute.
Everything free does, plus a Slack digest per workspace if you would rather read it there.
Where does it run?
npx pierre-review on your machine, or the hosted service. No repository limit, no user limit.
The same two places. Pro unlocks on whichever one you already use.
Pro lists at $25 per user, and is free for a month to anyone who asks for it. A user is someone who signs in — bots never count, and there is no per-repository charge.
What a free account actually sees. The paid panes are listed and locked rather than hidden, so you can tell what you are not buying before you buy it — and the free metrics above them are the ones that moved here off another tab.
Start free. Every repository, nothing missed.
Sign in with GitHub and the first repositories are on screen in a couple of minutes, or run the whole free tier on your own machine and keep the data there. For Pro, send a message — it is free for a month, for asking.
Card payments are not switched on yet, so Pro is not sold from this page — it is given, a month at a time, to anyone who asks. No card, nothing to cancel. Or run the free tier locally: npx pierre-review